Time Trends in the Use of Cholesterol-Lowering Agents in Older Adults:  The Cardiovascular Health Study FREE

RECOMMENDATIONS for the detection and control of hypercholesterolemia have been available since 1988, when the National Cholesterol Education Program (NCEP) guidelines were first issued.¹ The guidelines, together with the scientific evidence supporting them, have resulted in increased use of cholesterol-lowering drug therapy,^2- 4 although screening for hypercholesterolemia reportedly is incomplete^3- 5 and undertreatment among men and women at high risk of coronary heart disease (CHD) has been observed.^5- 8 While there is general agreement that middle-aged individuals with CHD should be treated for hypercholesterolemia, the benefit of treating older adults remains controversial.^9- 14 The 1993 NCEP guidelines recommend treating elderly patients who are otherwise in good health with cholesterol-lowering therapy.¹⁵ In contrast, a recent report from the American College of Physicians discourages screening men and women aged 75 years or older for hypercholesterolemia.¹⁶

The Cardiovascular Health Study (CHS), a cohort study designed to investigate risk factors for CHD in men and women aged 65 years or older,¹⁷ provided the opportunity to examine the impact of NCEP guidelines on physicians' prescribing patterns for their elderly patients. At baseline, in 1989-1990, 46% of CHS participants were eligible for lipoprotein analysis and 36% for dietary or drug therapy according to the 1988 NCEP guidelines.¹⁸ In the present study, we investigated the potential impact of the revised NCEP guidelines¹⁵ by examining the prevalence of treatment eligibility in the 5201 CHS participants initially enrolled in 1989-1990 and the 687 African American participants enrolled in 1992-1993. In addition, we used follow-up data to investigate the proportion of CHS participants who initiated cholesterol-lowering drug therapy during 6 years of follow-up; the use of the different classes of cholesterol-lowering drugs in each follow-up year; the prevalence of cholesterol-lowering drug therapy at the seventh (1995-1996) examination; and the factors that may have influenced the initiation of cholesterol-lowering drug therapy.

The methods have been described¹⁷ and are briefly summarized herein.

The CHS cohort initially consisted of 5201 noninstitutionalized men and women aged 65 years or older, who were recruited beginning in June 1989. The cohort was recruited from 4 US communities: Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Pittsburgh (Allegheny County), Pennsylvania. Beginning in June 1992, 687 African American adults were recruited into the cohort from Forsyth County, Sacramento County, and Pittsburgh. Participants were sampled randomly from the Medicare eligibility lists of the Health Care Financing Administration. The original cohort was followed up for 6 years, and the African American cohort for 3 years, during which the study participants received their usual care by their own physicians.

The baseline examination consisted of a home interview and clinical examination. Data on demographics, medical history, and personal habits were collected from standardized questionnaires. Information on cholesterol-lowering medication used in the preceding 2 weeks was collected directly from prescription bottles. The clinical examination included sitting blood pressure, anthropometric measurements, and venipuncture. Aliquots of plasma or serum were prepared, frozen at 0°C, and shipped weekly on dry ice to the Central Blood Analysis Laboratory (Colchester, Vt). Plasma lipid analyses were performed with an automated system (Olympus Demand System, Olympus Corp, Lake Success, NY) and included measurement of total cholesterol, high-density lipoprotein (HDL) cholesterol, and triglyceride levels, standardized according to the Centers for Disease Control and Prevention. Low-density lipoprotein (LDL) cholesterol levels were calculated according to a published equation.¹⁹ Participants and their physicians were informed of the results of the lipid analyses. Venipuncture and plasma lipid analyses were repeated in the third year of follow-up for the CHS participants initially enrolled in 1989-1990. Assessment of cholesterol-lowering medication use was repeated annually during the follow-up years for both the initial and the minority cohorts.

Confirmation of a participant's report of previous myocardial infarction, prevalent angina, previous stroke, previous transient ischemic attack, and prevalent peripheral vascular disease were sought from objective measures collected at entry and from the review of hospital and physicians' records.¹⁷ During the 6 years of follow-up, the investigation of potential incident events sometimes uncovered unreported prebaseline events.²⁰ This information was used to update baseline status for myocardial infarction, angina, claudication, stroke, and transient ischemic attack.

Prevalent CHD was defined as previous diagnosis of myocardial infarction, confirmed history of angina, confirmed previous coronary bypass surgery, confirmed angioplasty of the coronary arteries, reported carotid endarterectomy, or reported bypass procedure on a leg artery to match the CHD definition of the 1993 NCEP guidelines.¹⁵ Family history of premature CHD was defined as myocardial infarction in a sibling before the age of 55 years. Hypertension was defined as a seated systolic blood pressure greater than or equal to 160 mm Hg, or a seated diastolic blood pressure greater than or equal to 95 mm Hg, or prescribed antihypertensive medications together with a self-reported history of high blood pressure.

The CHS participants who were untreated at baseline were classified into 1 of 3 mutually exclusive categories of indicated treatment using an algorithm based on the 1993 NCEP guidelines.¹⁵ The categories were (1) ineligible for treatment; (2) eligible for dietary therapy; and (3) eligible for cholesterol-lowering drug therapy after a trial of dietary therapy. The classification was based on the presence of CHD, LDL cholesterol levels, total cholesterol levels, HDL cholesterol levels, and the presence of other risk factors including age (a positive risk factor for all members of the CHS cohort), diabetes, hypertension, current smoking, and family history of premature CHD.¹⁵

The classification into these categories of treatment eligibility represents a simplification of the NCEP guidelines. The guidelines, in fact, suggest individual judgment in deciding whether intensive cholesterol-lowering therapy should be undertaken in the elderly. In addition, the NCEP guidelines require the average of 2 cholesterol measurements, while a single measurement was used to classify treatment eligibility of the CHS participants. This single measurement may have resulted in some misclassification of individuals. Because participants and their physicians received the CHS cholesterol results, they had an opportunity for additional measurements on which treatment decisions could be based.

The 1988 NCEP guidelines were in effect at baseline and were updated during the follow-up in 1993. The 2 sets of guidelines were based on similar algorithms with a notable difference in the importance given to prevalent CHD (Table 1). In the updated guidelines, cholesterol-lowering drug therapy of patients with prevalent CHD is recommended at lower LDL cholesterol levels. We did the analyses both ways, using the 1993 or the 1988 NCEP guidelines to classify the CHS participants in categories of treatment eligibility. There was not much difference in the results (see the "Results" section). We chose the 1993 guidelines for the primary analysis presented in our study because we wanted to document cholesterol-lowering drug therapy in the elderly according to current recommendations.

We computed the proportion of CHS participants treated with cholesterol-lowering drugs in 1989-1990, the baseline years of the initial cohort (n = 5197 after excluding 4 participants with missing information); in 1992-1993, the baseline years of the African American cohort (n = 686 after excluding 1 participant with missing information); and again in 1995-1996, the last year of follow-up, including all survivors with drug information (n = 4529). The proportion of CHS participants treated with individual cholesterol-lowering drugs at baseline and in each follow-up year was also computed, including in a given year all surviving participants with drug information. To investigate the initiation of cholesterol-lowering drug therapy during the follow-up years as a function of recommended treatment and prevalent CHD, we selected the CHS participants who were untreated at baseline and were eligible either for dietary or drug therapy according to the 1993 NCEP guidelines (n = 2121 in 1989-1990, 1314 without CHD and 807 with CHD; and n = 203 in 1992-1993, 121 without CHD and 82 with CHD), and we computed the proportion treated with cholesterol-lowering drugs during the follow-up years, including in a given year all survivors with drug information.

To assess which factors might have influenced the initiation of cholesterol-lowering drug therapy, we performed Cox proportional hazards regression analyses²¹ using statistical software (SPSS for Windows, version 6.1, SPSS Inc, Chicago, Ill). Subjects were excluded if they were taking cholesterol-lowering drugs at baseline (n = 333); if they were not eligible for dietary or drug therapy according to the 1993 NCEP guidelines (n = 3111); if their eligibility for dietary or drug therapy could not be established because of missing laboratory values (n = 116); if data on drug therapy at baseline or 1 of the follow-up examinations were missing (n = 264); or if they were not followed up for at least 1 year (n = 92). These Cox proportional hazards regression analyses thus included 1972 CHS participants who were not taking cholesterol-lowering drugs at baseline and who were eligible for either drug therapy or dietary therapy, according to the 1993 NCEP guidelines. The average time to censoring was 4.3 years. During the follow-up years, 284 subjects started taking cholesterol-lowering drugs. Information was not collected on the exact date of starting drug therapy, only the regular use of drugs in the 2 weeks preceding each annual clinic visit was ascertained; thus, for patients who started taking cholesterol-lowering drugs between 2 clinic visits, the starting time was arbitrarily selected as the midpoint between the 2 relevant annual clinic visits. Consequently, there were 6 possible starting times during the 6 years of follow-up: 0.5, 1.5, 2.5, 3.5, 4.5, and 5.5 years.

The factors that predicted the initiation of cholesterol-lowering drug therapy were selected by stepwise procedures from among the following baseline variables: the continuous variables of age, total cholesterol level, LDL cholesterol level, HDL cholesterol level, triglyceride level, creatinine level, body weight, body mass index (measured as the weight in kilograms divided by the square of the height in meters), alcohol consumption (grams per week), number of CHD risk factors, number of instrumental activities of daily living performed with difficulty (range, 0-6), and date of entry in the CHS cohort; the binary variables of sex, prevalent CHD, diabetes, hypertension, current smoking, family history of premature CHD, and race (white vs nonwhite); and the categorical variables of education (<grade 12, high school or vocational school, and college graduate or professional education), family income (<$5000, $5000-$7999, $8000-$11,999, $12,000-$15,999, $16,000-$24,999, $25,000-$34,999, $35,000-$49,999, and ≥$50,000), and clinic site. Seven of these variables were included as time-dependent variables: total cholesterol, LDL cholesterol, and HDL cholesterol levels were updated once for members of the original cohort in the third year of follow-up when the results of new tests became available; prevalent CHD, diabetes, hypertension, and instrumental activities of daily living were updated in each year of follow-up. Prevalent CHD and diabetes in each year of follow-up were defined cumulatively as CHD and diabetes at any time during the previous study years. Hypertension and difficulties with instrumental activities of daily living were updated in each year of the study with the most recent data. Stepwise procedures that included the female subjects only were also performed including estrogen use as a time-dependent variable in addition to the variables described earlier.

The relative risks that are shown were obtained from models including only the predictors. Age was included in the final Cox proportional hazards regression analyses as a categorical variable. Relative risks for categories of total cholesterol level and HDL cholesterol level were computed from the regression coefficients for the continuous variables using the mean values of total cholesterol and HDL cholesterol levels of the participants in the different categories.

These analyses were based on the updated CHS databases that incorporate minor corrections up to March 1997.

Among the 5201 CHS participants enrolled in 1989-1990, about 5% were taking cholesterol-lowering drugs at baseline, 4.5% of men and 5.9% of women (Table 2). Among participants with a history of CHD, 8.2% of the men and 9.9% of the women were taking cholesterol-lowering drugs. Among participants untreated at baseline, 18.7% of men and 22.5% of women would have been eligible for drug therapy after a trial of dietary therapy, according to the 1993 NCEP guidelines. Another 21.7% of men and 22.7% of women would have been eligible for dietary therapy (Table 2).

The 1988 NCEP guidelines were in effect during the baseline years of the cohort and were updated during the follow-up years. Among participants untreated at baseline, 11.7% of men and 16.1% of women would have been eligible for drug therapy after a trial of dietary therapy according to the 1988 NCEP guidelines. A similar proportion of men and women would have been eligible for dietary therapy under either set of guidelines. We chose to base our analyses on the 1993 guidelines to evaluate treatment with reference to contemporary guidelines and because the conclusions of the analyses were similar using either set of guidelines.

Among the 687 African American participants enrolled in 1992-1993, 6.3% of men and 9.5% of women were taking cholesterol-lowering drugs at baseline (Table 3). Among untreated African American participants, 6.7% of men and 17.2% of women would have been eligible for drug therapy after a trial of dietary therapy (Table 3). During the same calendar years, 6.3% of men and 8.1% of women of the original cohort in the third year of follow-up were taking cholesterol-lowering drugs therapy and 13.0% of untreated men and 13.3% of untreated women were eligible for drug therapy.

The proportion of CHS participants untreated at baseline who started drug therapy during the follow-up years is illustrated in Figure 1, A and B. Among participants who were eligible for drug therapy after a trial of dietary therapy in 1989-1990, according to the 1993 NCEP guidelines, less than 20% initiated drug therapy in the 6 years of follow-up. Among participants eligible for drug therapy in 1989-1990 according to the 1988 NCEP guidelines, less than 25% initiated drug therapy in the 6 years of follow-up (data not shown). The proportion of participants who had initiated drug therapy after 6 years was identical in participants with CHD and in participants without CHD, 16.4% of survivors (Figure 1, A and B). Among minority participants enrolled in 1992-1993, the proportion of participants who had initiated drug therapy after 3 years was 8.7% among participants with CHD and eligible for drug therapy (Figure 1, A), and 5.4% among participants without CHD and eligible for drug therapy (Figure 1, B). Initiation of drug therapy was similar for men and women (data not shown). Initiation of drug therapy was lower for participants enrolled at the Washington County site (12% after 6 years, among participants eligible for drug therapy) than for participants at the other 3 CHS clinical sites (18% after 6 years among participants eligible for drug therapy).

In 1995-1996, the proportion of CHS participants taking cholesterol-lowering drugs was 8.1% of all men and 10.0% of all women. The proportion of CHS participants with prevalent CHD who were taking cholesterol-lowering drugs was increased from 8% to 10% in 1989-1990 (Table 2) to 16% in 1995-1996 (Table 4). Prevalence of cholesterol-lowering drug therapy decreased for those aged 78 years or older (Table 4). Prevalence of cholesterol-lowering drug therapy increased with total cholesterol levels, from 4% to 23% among participants free of CHD in the lowest and highest quintile of cholesterol levels, respectively, and from 12% to 29% among participants with prevalent CHD in the lowest and highest quintile, respectively (data not shown).

The 2 classes of cholesterol-lowering drugs that were used the most were 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors and fibric acid derivatives (Figure 1, C). There was a noticeable trend toward increased use of HMG-CoA reductase inhibitors during the follow-up years, from 1.9% of all participants in 1989-1990 to 7.5% in 1995-1996. This trend was observed at all 4 clinical sites (data not shown). The use of fibric acid derivatives declined from 1.9% in 1989-1990 to 1.4% in 1995-1996 and the use of bile acid sequestrants, nicotinic acid, and probucol declined over the 6 years from initial levels of less than 1% each (Figure 1, C).

We investigated which characteristics predicted the initiation of cholesterol-lowering drug therapy during follow-up years among CHS participants who were untreated at baseline and eligible either for drug or dietary therapy (Table 5). Seven characteristics were independently associated with the likelihood of starting cholesterol-lowering drug therapy: a history of CHD, total cholesterol levels, HDL cholesterol levels, hypertension, age, difficulties with activities of daily living, and clinical site. In analyses adjusted for other characteristics, untreated participants with a history of CHD were twice as likely to start cholesterol-lowering drug therapy as participants without prior CHD (Table 5). Elevated total cholesterol levels were strongly associated with the initiation of cholesterol-lowering drug therapy. Participants with total cholesterol levels above 6.87 mmol/L (266 mg/dL) were nearly 9 times as likely to start cholesterol-lowering drug therapy than participants with cholesterol levels below 5.40 mmol/L (209 mg/dL). When taking into account total cholesterol levels, LDL cholesterol levels had no further relationship with the initiation of cholesterol-lowering drug therapy. However, the HDL cholesterol level was independently associated with the likelihood of initiating cholesterol-lowering drug therapy. Participants with an HDL cholesterol level of 1.55 mmol/L (60 mg/dL) or higher were half as likely to start cholesterol-lowering drug therapy than participants with an HDL cholesterol level below 0.90 mmol/L (35 mg/dL).

Age was associated with the initiation of cholesterol-lowering drug therapy. Participants younger than 70 years in the baseline years were 4 times more likely to start cholesterol-lowering drug therapy during the follow-up years than participants 80 years or older (Table 5). In these multivariate analyses, a history of CHD did not modify the relationship of age to treatment (data not shown). A report of difficulties with activities of daily living was associated with the initiation of cholesterol-lowering drug therapy. An increase of 1 task performed with difficulty was associated with a 25% decrease in the likelihood of initiating cholesterol-lowering drug therapy (Table 5). Finally, participants enrolled at the Washington County site were less likely to start cholesterol-lowering drug therapy than participants at the other sites (Table 5).

The characteristics associated with the initiation of drug therapy were similar for participants eligible for drug therapy and for participants eligible for dietary therapy (Table 5). In stepwise procedures stratified on recommended treatment, no additional characteristics were selected for either group of recommended treatment (data not shown). Other characteristics that also should be considered according to the NCEP guidelines—diabetes, smoking, family history of premature CHD, and total number of risk factors—were not associated with the initiation of cholesterol-lowering drug therapy in univariate analyses (not shown) and were not selected during the stepwise procedures. Additional characteristics including sex, race, education, income at baseline, weight, body mass index, alcohol use, baseline creatinine level, baseline triglyceride level, and time of entry in the cohort also were not selected.

About 19% (399) of all male and 23% (626) of all female CHS participants who were untreated at baseline would have been eligible for cholesterol-lowering drug therapy after a trial of dietary therapy according to the 1993 NCEP guidelines. Yet among the 1025 participants eligible for drug therapy, less than 20% of the survivors were actually taking cholesterol-lowering drugs in 1995-1996 at the end of 6 years of follow-up.

We found no differences in the extent of cholesterol-lowering drug therapy between the African American cohort and the original, mostly white cohort. Similar proportions of the men and women in both cohorts were treated with cholesterol-lowering drug therapy during the same calendar years, and among the untreated men and women, similar proportions were eligible for drug therapy according to the 1993 NCEP guidelines. Furthermore, in both univariate and multivariate analyses, race was not associated with the initiation of cholesterol-lowering drug therapy during the follow-up years. African Americans have been reported to be less likely to be treated for hypercholesterolemia.^3,5,8 In a cross-sectional analysis of the Heart and Estrogen/Progestin Replacement Study (HERS), a study of postmenopausal women aged 44 to 79 years, all of whom have CHD, African American race was reported to predict the lack of treatment with cholesterol-lowering drug therapy.⁸ It is possible that the difference in treatment between whites and African Americans is eliminated in the elderly, when the prevalence of treatment is much lower for all.

The cholesterol-lowering drugs that were used most often in the CHS cohort in 1989-1990 were HMG-CoA reductase inhibitors and fibric acid derivatives. During the 6 years of follow-up, we observed a dramatic increase in the use of HMG-CoA reductase inhibitors, from 2% of all participants in 1989-1990 to 7.5% in 1995-1996. Similar trends of increased prescriptions for HMG-CoA reductase inhibitors in the United States suggest that the choice of drugs for hypercholesterolemia does not differ greatly by age.^22- 23 Because HMG-CoA reductase inhibitors are well tolerated,²⁴ the increase in HMG-CoA reductase inhibitor therapy may reflect an increase in compliance, as well as an increase in HMG-CoA reductase inhibitor prescriptions. While the effect of HMG-CoA reductase inhibitors on CHD incidence had not yet been established during the early years of the CHS cohort follow-up, 3 clinical trials have now demonstrated a decrease in coronary events and mortality in subjects treated with HMG-CoA reductase inhibitors used both in primary and secondary prevention.^25- 27 Meta-analyses have also shown a reduction in stroke in subjects treated with HMG-CoA reductase inhibitors used in secondary prevention.^28- 29 Although no clinical trials have yet specifically targeted the elderly, HMG-CoA reductase inhibitors are effective at lowering cholesterol through age 75 years with a single dose per day.^30- 33

The 1993 NCEP recommendations on cholesterol-lowering drug therapy are based on total and LDL cholesterol levels, prevalent CHD, and the presence of other risk factors for CHD.¹⁵ We found that only prevalent CHD, total cholesterol levels, HDL cholesterol levels, and hypertension were associated with the initiation of cholesterol-lowering drug therapy among the CHS participants. Other risk factors listed in the 1993 guidelines—diabetes, smoking, family history of premature CHD, total number of risk factors, and body mass index (listed in the 1988 guidelines)—were not associated with the initiation of cholesterol-lowering drug therapy. While most physicians participating in a 1990 survey reported using the NCEP recommendations,² our study results suggest that the recommendations are only partially followed.

Undertreatment with cholesterol-lowering drug therapy, relative to the NCEP guidelines, has been reported in all age groups.^5- 8 Among the postmenopausal women with CHD in the HERS study, 47% were reported to receive cholesterol-lowering drug therapy.⁸ In contrast, among the female CHS participants with CHD, 10% at baseline and 16% in 1995-1996 were taking cholesterol-lowering medication. One reason for the difference in treatment might be a difference in treatment eligibility of the untreated subjects in the 2 studies. Information on the LDL cholesterol levels of the untreated subjects is not available in the HERS report; however, the distributions of other CHD risk factors, such as smoking, diabetes, hypertension, and body mass index, appeared similar in the HERS participants and the CHS cohort women with CHD, and differences in treatment eligibility are not likely to explain the 3- to 4-fold difference in cholesterol-lowering drug therapy. Another possibility is that treatment is less prevalent among the older elderly. In support of this interpretation, we observed that the use of cholesterol-lowering drugs in 1995-1996 decreased with increasing age and that the likelihood of initiating cholesterol-lowering drug therapy during the 6 years of follow-up decreased with increasing age independent of other risk factors. A third possibility is that volunteers for a clinical trial such as HERS may differ from a general population sample such as the CHS participants.

The 1993 NCEP guidelines indicate that cholesterol-lowering therapy is not suitable for many elderly patients because of competing illnesses.¹⁵ In agreement with these recommendations, the use of cholesterol-lowering drug therapy among the CHS participants decreased with increasing disability. However, the CHS participants appeared relatively healthy. Persons who were institutionalized, wheelchair-bound in the home, or receiving hospice treatment, radiation therapy, or chemotherapy for cancer at baseline were excluded from the cohort.¹⁷ As a result, less than 5% of the participants reported poor health at baseline and poor health is not a likely explanation for the low levels of treatment in the cohort as a whole. The low levels of treatment may reflect the uncertainty about the risk-benefit of cholesterol-lowering therapy in the elderly.^9- 14 Whether there is an age at which treatment is no longer beneficial is simply not known.³⁴ The benefits of simvastatin treatment observed in the Scandinavian Simvastatin Survival Study applied to patients 60 to 70 years, up to 75 years at the end of the trial, as well as to younger patients.^25,35 In addition, a meta-analysis indicates a beneficial effect of pravastatin sodium on disease events in secondary prevention through age 75 years.²⁸ Recent evidence suggests that elevated levels of cholesterol remain a risk factor for CHD death in the elderly in otherwise good health.³⁶ A clinical trial will be needed to confirm the inference that treatment may be beneficial after age 75 years.

The strengths of our study include the prospective study design, the reliable ascertainment of CHD events, and the richness of the information collected from the study participants in each study year. The main limitation of this study is that a single measurement of cholesterol was used to classify the participants into categories of treatment eligibility. This single measurement may have misclassified a few individuals; however, regression to the mean cannot account for the low levels of treatment in the sample as a whole because the CHS sample of participants was not selected for elevated levels of cholesterol. Another limitation is the lack of information on dietary therapy. Older patients and their physicians may have preferred dietary therapy over pharmacological treatment, and dietary therapy might explain a small proportion of untreated participants eligible for therapy. Finally, we had no information on compliance. We do not know if study participants did not fill prescriptions for cholesterol-lowering drugs that had been written by their physicians and to what extent the lack of treatment reflected patients' choice. However, income was not associated with the initiation of cholesterol-lowering drug therapy.

While prevalent CHD was associated with increased likelihood of initiation of cholesterol-lowering drug therapy during the follow-up years by about 2-fold, fewer than 20% of participants with prevalent CHD, untreated at baseline and for whom the 1993 NCEP guidelines suggested drug therapy after a trial of dietary therapy, initiated drug therapy during follow-up. In view of the clinical trial evidence for benefit, patients aged 65 to 75 years and with a history of CHD appear undertreated with cholesterol-lowering drug therapy.

Accepted for publication February 12, 1998.

This study was supported by contracts N01 HC 85079, N01 HC 85080, N01 HC 85081, N01 HC 85082, N01 HC 85083, N01 HC 85084, N01 HC 85085, and N01 HC 85086 from the National Heart, Lung, and Blood Institute, Bethesda, Md, and grant AC 09556 from the National Institute on Aging, Bethesda.

Dr Psaty is a Merck/SER Clinical Epidemiology Fellow, sponsored by the Merck Co Foundation, Rahway, NJ, and the Society for Epidemiologic Research, Baltimore, Md.

We appreciate the comments, criticisms, and suggestions that Susan R. Heckbert, MD, PhD, provided on early drafts of this article, and the stimulating discussions about the statistical analyses and assistance with data management of Nick L. Smith, PhD.

Reprints: Rozenn N. Lemaitre, PhD, MPH, Cardiovascular Health Research Unit, Metropolitan Park, East Tower, Suite 1360, 1730 Minor Ave, Seattle, WA 98101.